unrag 0.2.5 → 0.2.7

package/registry/extractors/video-transcribe/index.ts

@@ -1,7 +1,21 @@
-import { experimental_transcribe as transcribe } from "ai";
-import type { AssetExtractor } from "../../core/types";
+import { experimental_transcribe as transcribe, type TranscriptionModel } from "ai";
+import type { AssetExtractor, ExtractedTextItem } from "../../core/types";
 import { getAssetBytes } from "../_shared/fetch";
 
+/**
+ * Model reference type that accepts both string gateway IDs and TranscriptionModel instances.
+ */
+type TranscriptionModelRef = string | TranscriptionModel;
+
+/**
+ * Transcription segment from the AI SDK.
+ */
+interface TranscriptionSegment {
+  text?: string;
+  startSecond?: number;
+  endSecond?: number;
+}
+
 /**
  * Video transcription by sending the video file to the AI SDK transcription API.
  *
@@ -28,43 +42,45 @@ export function createVideoTranscribeExtractor(): AssetExtractor {
       const abortSignal = AbortSignal.timeout(cfg.timeoutMs);
 
       const result = await transcribe({
-        model: cfg.model as any,
-        audio: bytes as any,
+        model: cfg.model as TranscriptionModelRef,
+        audio: bytes,
         abortSignal,
       });
 
-      const segments: any[] = Array.isArray((result as any)?.segments)
-        ? (result as any).segments
+      const segments: TranscriptionSegment[] = Array.isArray(result.segments)
+        ? result.segments
         : [];
 
       if (segments.length > 0) {
+        const textItems: ExtractedTextItem[] = segments
+          .map((s, i) => {
+            const t = String(s?.text ?? "").trim();
+            if (!t) return null;
+            const start = Number(s?.startSecond ?? NaN);
+            const end = Number(s?.endSecond ?? NaN);
+            return {
+              label: `segment-${i + 1}`,
+              content: t,
+              ...(Number.isFinite(start) && Number.isFinite(end)
+                ? { timeRangeSec: [start, end] as [number, number] }
+                : {}),
+            };
+          })
+          .filter((item): item is ExtractedTextItem => item !== null);
+
         return {
-          texts: segments
-            .map((s, i) => {
-              const t = String(s?.text ?? "").trim();
-              if (!t) return null;
-              const start = Number(s?.startSecond ?? NaN);
-              const end = Number(s?.endSecond ?? NaN);
-              return {
-                label: `segment-${i + 1}`,
-                content: t,
-                ...(Number.isFinite(start) && Number.isFinite(end)
-                  ? { timeRangeSec: [start, end] as [number, number] }
-                  : {}),
-              };
-            })
-            .filter(Boolean) as any,
+          texts: textItems,
           diagnostics: {
             model: cfg.model,
             seconds:
-              typeof (result as any)?.durationInSeconds === "number"
-                ? (result as any).durationInSeconds
+              typeof result.durationInSeconds === "number"
+                ? result.durationInSeconds
                 : undefined,
           },
         };
       }
 
-      const text = String((result as any)?.text ?? "").trim();
+      const text = (result.text ?? "").trim();
       if (!text) return { texts: [], diagnostics: { model: cfg.model } };
 
       return {

package/registry/manifest.json

@@ -0,0 +1,346 @@
+{
+  "version": 1,
+  "extractors": [
+    {
+      "id": "pdf-text-layer",
+      "extractorName": "pdf:text-layer",
+      "group": "PDF",
+      "label": "pdf-text-layer",
+      "description": "Fast/cheap extraction via PDF text layer",
+      "hint": "recommended",
+      "defaultSelected": true,
+      "workerOnly": false,
+      "configComplexity": "needs-dep",
+      "fileTypes": ["pdf"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "text",
+      "docsPath": "/docs/extractors/pdf/text-layer",
+      "deps": { "pdfjs-dist": "^5.4.149" },
+      "devDeps": {},
+      "factory": "createPdfTextLayerExtractor",
+      "assetProcessingFlagKeys": ["pdf_textLayer"]
+    },
+    {
+      "id": "pdf-llm",
+      "extractorName": "pdf:llm",
+      "group": "PDF",
+      "label": "pdf-llm",
+      "description": "LLM-based PDF extraction; higher cost",
+      "defaultSelected": false,
+      "workerOnly": false,
+      "configComplexity": "needs-api-key",
+      "fileTypes": ["pdf"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "text (markdown)",
+      "docsPath": "/docs/extractors/pdf/llm",
+      "deps": { "ai": "^6.0.3" },
+      "devDeps": {},
+      "factory": "createPdfLlmExtractor",
+      "assetProcessingFlagKeys": ["pdf_llmExtraction"]
+    },
+    {
+      "id": "pdf-ocr",
+      "extractorName": "pdf:ocr",
+      "group": "PDF",
+      "label": "pdf-ocr",
+      "description": "OCR scanned PDFs; requires native binaries",
+      "hint": "worker-only",
+      "defaultSelected": false,
+      "workerOnly": true,
+      "configComplexity": "advanced",
+      "fileTypes": ["pdf"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "text",
+      "docsPath": "/docs/extractors/pdf/ocr",
+      "deps": {},
+      "devDeps": {},
+      "factory": "createPdfOcrExtractor",
+      "assetProcessingFlagKeys": ["pdf_ocr"]
+    },
+    {
+      "id": "image-ocr",
+      "extractorName": "image:ocr",
+      "group": "Image",
+      "label": "image-ocr",
+      "description": "Extract text from images via vision LLM",
+      "defaultSelected": false,
+      "workerOnly": false,
+      "configComplexity": "needs-api-key",
+      "fileTypes": ["jpg", "png", "webp", "gif"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "text",
+      "docsPath": "/docs/extractors/image/ocr",
+      "deps": { "ai": "^6.0.0" },
+      "devDeps": {},
+      "factory": "createImageOcrExtractor",
+      "assetProcessingFlagKeys": ["image_ocr"]
+    },
+    {
+      "id": "image-caption-llm",
+      "extractorName": "image:caption-llm",
+      "group": "Image",
+      "label": "image-caption-llm",
+      "description": "Generate captions for images via vision LLM",
+      "defaultSelected": false,
+      "workerOnly": false,
+      "configComplexity": "needs-api-key",
+      "fileTypes": ["jpg", "png", "webp", "gif"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "caption",
+      "docsPath": "/docs/extractors/image/caption-llm",
+      "deps": { "ai": "^6.0.0" },
+      "devDeps": {},
+      "factory": "createImageCaptionLlmExtractor",
+      "assetProcessingFlagKeys": ["image_captionLlm"]
+    },
+    {
+      "id": "audio-transcribe",
+      "extractorName": "audio:transcribe",
+      "group": "Audio",
+      "label": "audio-transcribe",
+      "description": "Speech-to-text transcription",
+      "defaultSelected": false,
+      "workerOnly": false,
+      "configComplexity": "needs-api-key",
+      "fileTypes": ["mp3", "wav", "ogg", "m4a"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "transcript",
+      "docsPath": "/docs/extractors/audio/transcribe",
+      "deps": { "ai": "^6.0.0" },
+      "devDeps": {},
+      "factory": "createAudioTranscribeExtractor",
+      "assetProcessingFlagKeys": ["audio_transcription"]
+    },
+    {
+      "id": "video-transcribe",
+      "extractorName": "video:transcribe",
+      "group": "Video",
+      "label": "video-transcribe",
+      "description": "Transcribe video audio track",
+      "defaultSelected": false,
+      "workerOnly": false,
+      "configComplexity": "needs-api-key",
+      "fileTypes": ["mp4", "webm", "mov"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "transcript",
+      "docsPath": "/docs/extractors/video/transcribe",
+      "deps": { "ai": "^6.0.0" },
+      "devDeps": {},
+      "factory": "createVideoTranscribeExtractor",
+      "assetProcessingFlagKeys": ["video_transcription"]
+    },
+    {
+      "id": "video-frames",
+      "extractorName": "video:frames",
+      "group": "Video",
+      "label": "video-frames",
+      "description": "Sample frames + analyze via vision LLM; requires ffmpeg",
+      "hint": "worker-only",
+      "defaultSelected": false,
+      "workerOnly": true,
+      "configComplexity": "advanced",
+      "fileTypes": ["mp4", "webm", "mov"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "frame descriptions",
+      "docsPath": "/docs/extractors/video/frames",
+      "deps": { "ai": "^6.0.0" },
+      "devDeps": {},
+      "factory": "createVideoFramesExtractor",
+      "assetProcessingFlagKeys": ["video_frames"]
+    },
+    {
+      "id": "file-text",
+      "extractorName": "file:text",
+      "group": "Files",
+      "label": "file-text",
+      "description": "Extract text/markdown/json/html from common text files",
+      "hint": "recommended",
+      "defaultSelected": true,
+      "workerOnly": false,
+      "configComplexity": "zero-config",
+      "fileTypes": ["txt", "md", "json", "csv"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "text",
+      "docsPath": "/docs/extractors/file/text",
+      "deps": {},
+      "devDeps": {},
+      "factory": "createFileTextExtractor",
+      "assetProcessingFlagKeys": ["file_text"]
+    },
+    {
+      "id": "file-docx",
+      "extractorName": "file:docx",
+      "group": "Files",
+      "label": "file-docx",
+      "description": "Extract text from .docx files",
+      "defaultSelected": false,
+      "workerOnly": false,
+      "configComplexity": "needs-dep",
+      "fileTypes": ["docx"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "text",
+      "docsPath": "/docs/extractors/file/docx",
+      "deps": { "mammoth": "^1.10.0" },
+      "devDeps": {},
+      "factory": "createFileDocxExtractor",
+      "assetProcessingFlagKeys": ["file_docx"]
+    },
+    {
+      "id": "file-pptx",
+      "extractorName": "file:pptx",
+      "group": "Files",
+      "label": "file-pptx",
+      "description": "Extract text from .pptx slides",
+      "defaultSelected": false,
+      "workerOnly": false,
+      "configComplexity": "needs-dep",
+      "fileTypes": ["pptx"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "text",
+      "docsPath": "/docs/extractors/file/pptx",
+      "deps": { "jszip": "^3.10.1" },
+      "devDeps": {},
+      "factory": "createFilePptxExtractor",
+      "assetProcessingFlagKeys": ["file_pptx"]
+    },
+    {
+      "id": "file-xlsx",
+      "extractorName": "file:xlsx",
+      "group": "Files",
+      "label": "file-xlsx",
+      "description": "Extract tables from .xlsx spreadsheets",
+      "defaultSelected": false,
+      "workerOnly": false,
+      "configComplexity": "needs-dep",
+      "fileTypes": ["xlsx"],
+      "inputModes": ["file", "url", "buffer"],
+      "output": "text (csv)",
+      "docsPath": "/docs/extractors/file/xlsx",
+      "deps": { "xlsx": "^0.18.5" },
+      "devDeps": {},
+      "factory": "createFileXlsxExtractor",
+      "assetProcessingFlagKeys": ["file_xlsx"]
+    }
+  ],
+  "connectors": [
+    {
+      "id": "notion",
+      "displayName": "Notion",
+      "types": ["docs", "wiki", "db"],
+      "description": "Sync pages, databases, and blocks from Notion workspaces",
+      "status": "available",
+      "docsPath": "/docs/connectors/notion",
+      "deps": { "@notionhq/client": "^2.2.16" },
+      "devDeps": {},
+      "envVars": [
+        { "name": "NOTION_TOKEN", "required": true, "notes": "Server-only Notion integration token." }
+      ]
+    },
+    {
+      "id": "google-drive",
+      "displayName": "Google Drive",
+      "types": ["files", "docs"],
+      "description": "Ingest Docs/Sheets exports and shared folders",
+      "status": "available",
+      "docsPath": "/docs/connectors/google-drive",
+      "deps": { "googleapis": "^148.0.0", "google-auth-library": "^10.0.0" },
+      "devDeps": {},
+      "envVars": [
+        {
+          "name": "GOOGLE_SERVICE_ACCOUNT_JSON",
+          "required": false,
+          "notes": "Service account JSON credentials (server-only)."
+        },
+        { "name": "GOOGLE_CLIENT_ID", "required": false, "notes": "OAuth client id (server-only)." },
+        {
+          "name": "GOOGLE_CLIENT_SECRET",
+          "required": false,
+          "notes": "OAuth client secret (server-only)."
+        },
+        { "name": "GOOGLE_REDIRECT_URI", "required": false, "notes": "OAuth redirect URI." }
+      ]
+    },
+    {
+      "id": "github",
+      "displayName": "GitHub",
+      "types": ["code", "docs"],
+      "description": "Ingest repositories (Markdown, READMEs, docs folders) and issues/PRs",
+      "status": "coming-soon",
+      "docsPath": null,
+      "deps": {},
+      "devDeps": {}
+    },
+    {
+      "id": "gitlab",
+      "displayName": "GitLab",
+      "types": ["code", "docs"],
+      "description": "Ingest repos + wiki pages for self-hosted documentation",
+      "status": "coming-soon",
+      "docsPath": null,
+      "deps": {},
+      "devDeps": {}
+    },
+    {
+      "id": "slack",
+      "displayName": "Slack",
+      "types": ["chat"],
+      "description": "Ingest channels (messages + threads) as searchable knowledge",
+      "status": "coming-soon",
+      "docsPath": null,
+      "deps": {},
+      "devDeps": {}
+    },
+    {
+      "id": "discord",
+      "displayName": "Discord",
+      "types": ["chat"],
+      "description": "Ingest server channels and threads for community support knowledge",
+      "status": "coming-soon",
+      "docsPath": null,
+      "deps": {},
+      "devDeps": {}
+    },
+    {
+      "id": "linear",
+      "displayName": "Linear",
+      "types": ["issues", "project"],
+      "description": "Ingest issues and project updates for internal knowledge",
+      "status": "coming-soon",
+      "docsPath": null,
+      "deps": {},
+      "devDeps": {}
+    },
+    {
+      "id": "dropbox",
+      "displayName": "Dropbox",
+      "types": ["files"],
+      "description": "Ingest shared folders and exported docs/files",
+      "status": "coming-soon",
+      "docsPath": null,
+      "deps": {},
+      "devDeps": {}
+    },
+    {
+      "id": "onedrive",
+      "displayName": "OneDrive",
+      "types": ["files"],
+      "description": "Ingest files and Office exports from Microsoft 365",
+      "status": "coming-soon",
+      "docsPath": null,
+      "deps": {},
+      "devDeps": {}
+    },
+    {
+      "id": "teams",
+      "displayName": "Microsoft Teams",
+      "types": ["chat"],
+      "description": "Ingest channels and conversations for internal support knowledge",
+      "status": "coming-soon",
+      "docsPath": null,
+      "deps": {},
+      "devDeps": {}
+    }
+  ]
+}
+
+

package/registry/store/drizzle-postgres-pgvector/store.ts

@@ -1,9 +1,26 @@
 import { documents, chunks, embeddings } from "./schema";
 import type { Chunk, VectorStore } from "../../core/types";
 import { eq, like, sql, type SQL } from "drizzle-orm";
-import type { PgDatabase } from "drizzle-orm/pg-core";
-
-type DrizzleDb = PgDatabase<any, any, any>;
+import type { PgDatabase, PgQueryResultHKT } from "drizzle-orm/pg-core";
+
+/**
+ * Accepts any Drizzle Postgres database instance regardless of schema type.
+ */
+type DrizzleDb = PgDatabase<PgQueryResultHKT, Record<string, unknown>>;
+
+/**
+ * Query row type for vector similarity search results.
+ */
+interface QueryRow {
+  id: string;
+  document_id: string;
+  source_id: string;
+  idx: number;
+  content: string;
+  token_count: number;
+  metadata: Record<string, unknown> | null;
+  score: number;
+}
 
 const sanitizeMetadata = (metadata: unknown) => {
   if (metadata === undefined) {
@@ -113,7 +130,7 @@ export const createDrizzleVectorStore = (db: DrizzleDb): VectorStore => ({
 
     const vectorLiteral = `[${embedding.join(",")}]`;
 
-    const rows = await db.execute(
+    const result = await db.execute(
       sql`
         select
           c.id,
@@ -133,7 +150,11 @@ export const createDrizzleVectorStore = (db: DrizzleDb): VectorStore => ({
       `
     );
 
-    return (rows as Array<Record<string, unknown>>).map((row) => ({
+    const rows: QueryRow[] = Array.isArray(result)
+      ? (result as QueryRow[])
+      : ((result as { rows?: QueryRow[] }).rows ?? []);
+
+    return rows.map((row) => ({
       id: String(row.id),
       documentId: String(row.document_id),
       sourceId: String(row.source_id),
@@ -157,4 +178,3 @@ export const createDrizzleVectorStore = (db: DrizzleDb): VectorStore => ({
   },
 });
 
-